Stroke is a debilitating neurological disease that can cause permanent motor disabilities, cognitive impairment, and death. Subcortical white matter (WM) stroke is the second leading cause of dementia. Neurovascular inflammation is closely associated with stroke, as both a cause and effect, but the specific mechanisms of post-stroke inflammation on tissue recovery are not well-understood. Prior research has shown that recognized cerebrovascular risk factors induce a neurovascular inflammatory response in mice and humans, driven in part by the chemokine C-X-C motif chemokine ligand 5 (CXCL5). This project studied the role of brain endothelial CXCL5 overexpression (OE) on post-WM stroke recovery. The results demonstrated a biphasic post-stroke effect of endothelial CXCL5 OE: initial increases in microglial response and ischemic damage, but a delayed improvement in WM recovery. Analysis of the microglial transcriptome in vitro supported an endothelial CXCL5-driven increase in microglial phagocytosis, allowing enhanced debris clearance and quicker resolution of post-stroke inflammation. This unique microglial response was associated with transcription factor SPI1. At the early recovery timepoint, CXCL5 OE drove increases in total SPI1 fluorescence intensity and number of SPI1-expressing microglia in the peri-infarct WM. Further clarifying this post-stroke inflammatory pathway could contribute to the development of therapeutics that leverage this neuroprotective response, thereby enhancing brain tissue repair after stroke.

Social interactions involve dynamic feedback that shapes decision-making, cooperation, and competition. Interbrain coupling, defined as the synchronized neural activity that emerges between individuals during social interaction, is a well-established phenomenon in biological systems and predicts future social behaviors. While artificial intelligence (AI) has been widely used to model individual cognition, it remains unclear whether independently trained AI agents can develop interbrain-like neural synchronization through interaction. Here, we examine the emergence of interbrain-like synchronization in multi-agent reinforcement learning systems. Long short-term memory (LSTM)-based agents are analyzed in cooperative (Gold Mining) and competitive (Predator-Prey) tasks. In cooperative settings, agents exhibit a transition from individual to coordinated strategies, accompanied by increased representational dimensionality and stronger inter-agent synchrony, particularly during cooperative events. In competitive settings, synchrony is selectively elevated between agents of opposing roles but remains low among agents with the same roles. To further characterize these dynamics, linear dynamical systems models are utilized to identify discrete latent states associated with social interaction events. Together, these findings demonstrate that independently trained artificial agents can develop interbrain-like synchronization alongside emergent social behavior, providing a computational framework for studying social neural dynamics.

The human brain contains incredible cellular diversity and specialization, with distinct neuronal and glial populations underlying complex cognition. Language, a uniquely human ability, relies on coordinated activity across brain regions. However, the transcriptional organization of these systems, particularly within glial populations, remains incompletely understood. Defining the cell-type organization of language-associated regions will illuminate how precise gene regulation supports advanced cognition. Postmortem tissue from nine donors each sampled across nine brain regions was processed for single-nucleus RNA sequencing utilizing 10x Genomics 3' kits. Reads were aligned to the human reference genome; low-quality nuclei and doublets were removed. Seurat was then used for unsupervised clustering with Harmony batch correction. Neuronal subtypes are known to diverge between cortex and basal ganglia; however, prior work has largely documented proportional rather than molecular subtype differences in glia. Here, we uncover novel molecular diversity within the oligodendrocyte lineage. BA4 (motor cortex) displays the greatest regional uniqueness; BA44, a Broca's area subdivision involved in speech motor planning, shares these molecular features. Distinct subpopulations are also identified in the putamen, globus pallidus internus, and brainstem. These findings suggest that oligodendrocyte diversity in the human brain is substantially underappreciated, with implications for the cellular basis of human language and cognition.

The orbital frontal cortex (OFC), and its connectivity with key limbic and reward regions, is vital to social behavior. However, how these circuitry dynamics are differentially engaged across neurodevelopmental populations with characteristically impacted social behavior, such as autism spectrum disorder (ASD), remains unclear. This study investigates OFC functional connectivity (FC) in relation to social responsiveness in ASD and typically developing (TD) controls. We performed seed-based analysis on resting-state fMRI (rsfMRI) data from 106 ASD and 92 TD participants (7-18 years). We extracted lateralized OFC, amygdala, nucleus accumbens (NAc), cingulate, hippocampus, insula, and thalamus as regions of interest from the Harvard-Oxford atlas. The Social Responsiveness Scale (SRS) subscales measured social traits such as reciprocal communication and social motivation. We then performed linear mixed-effects regression, correcting for multiple comparisons. We found a significant Group SRS Motivation subscale interaction on left OFC and NAc FC (q = 0.017, = -0.26) with no significant main effect of Group. Our results demonstrate that left OFC-NAc connectivity is weaker in ASD than TD even at comparable levels of social motivation. These findings suggest that atypical OFC-NAc connectivity emerges specifically in relation to social motivation, highlighting the importance of examining brain circuits in the context of behavioral dimensions.

There is an understanding gap in how dopaminergic imbalance alters decision strategies in Attention-Deficit/Hyperactivity Disorder (ADHD). Mice with heterozygous dopamine-transporters (DAT+/−) elevating extracellular dopamine was used. This produces hyperlocomotion, impulsivity, and learning/memory deficits, paralleling symptoms of ADHD. Does altered dopaminergic signaling lead to measurable variability in evidence accumulation? DAT+/− and WT littermates (5–6 weeks old; balanced sexes) were water-restricted; citric acid water given on non-training days. Behavioral sessions (~1hr) occurred in sound-attenuating chambers. Each trial began with a center port nose-poke, triggering a 1s sequence (4–20 flashes/s). A subsequent “go” cue signaled choice: right port for high rate (>12 flashes/s), left port for low (<12); 12 flashes/s served as an ambiguous reference. Premature withdrawals triggered a 1s timeout; correct choices yielded 5–7μL water. Wait duration increased from 0 to 1000ms over 10–12 weeks; training completed at ≥80% accuracy with full 1s wait. Performance was fit with a four-parameter logistic psychometric function: P(choose High) = γ + (1−λ−γ) F((x−α​)/β). Where x = flash rate; α = bias; β = sensitivity; γ = guess; λ = lapse. WT and DAT+/− mice acquired stable task performance consistent with prior studies. In DAT+/− mice β remained comparable, λ and γ rates increased, reflecting attentional instability rather than perceptual deficits. α had notable 50% threshold difference between high and low-rate choice.